[Paper Review] Predictive Analysis for Detection of Human Neck Postures using a robust integration of kinetics and kinematics

Korupalli V Rajesh Kumar, Susan Elias|arXiv (Cornell University)|Mar 12, 2020

Ergonomics and Musculoskeletal Disorders27 references4 citations

TL;DR

This paper proposes a smart neck-band system that integrates kinematic data from an IMU with kinetic data from OpenSim-simulated musculoskeletal models to predict human neck postures with 100% accuracy using Random Forest machine learning. The method fuses inertial sensor data with subject-specific neck musculoskeletal modeling to enable real-time detection of improper postures for preventive healthcare applications.

ABSTRACT

Human neck postures and movements need to be monitored, measured, quantified and analyzed, as a preventive measure in healthcare applications. Improper neck postures are an increasing source of neck musculoskeletal disorders, requiring therapy and rehabilitation. The motivation for the research presented in this paper was the need to develop a notification mechanism for improper neck usage. Kinematic data captured by sensors have limitations in accurately classifying the neck postures. Hence, we propose an integrated use of kinematic and kinetic data to efficiently classify neck postures. Using machine learning algorithms we obtained 100% accuracy in the predictive analysis of this data. The research analysis and discussions show that the kinetic data of the Hyoid muscles can accurately detect the neck posture given the corresponding kinematic data captured by the neck-band. The proposed robust platform for the integration of kinematic and kinetic data has enabled the design of a smart neck-band for the prevention of neck musculoskeletal disorders.

Motivation & Objective

• To develop a robust system for real-time detection of human neck postures to prevent cervical musculoskeletal disorders.
• To overcome limitations of kinematic-only sensing by integrating kinetic data from muscle forces for improved posture classification accuracy.
• To design a wearable smart neck-band that provides timely alerts for improper neck usage.
• To validate the effectiveness of combining IMU-derived kinematics with OpenSim-simulated kinetics for posture prediction.
• To achieve high-accuracy predictive analysis using machine learning on fused kinematic and kinetic data.

Proposed method

• An IMU-equipped elastic neck-band captures real-time kinematic data (acceleration, position, velocity) from subjects during neck movement tasks.
• Kinematic data from the neck-band's IMU is mapped to a subject-specific OpenSim musculoskeletal model using a functional calibration method with markers at key anatomical landmarks.
• The Inverse Kinematics (IK) tool in OpenSim processes the marker data to generate a motion file (.mot) representing joint kinematics.
• The Computed Muscle Control (CMC) tool in OpenSim uses the .mot file to compute muscle forces, activations, and tendon forces, generating kinetic data for neck muscles—particularly the Hyoid muscles.
• Kinematic and kinetic data (acceleration, position, tendon force) are extracted and used as input features for machine learning classification.
• A Random Forest classifier is trained and tested on the fused data to predict nine distinct neck postures with high accuracy.

Experimental results

Research questions

• RQ1 Can the integration of kinematic and kinetic data improve the accuracy of human neck posture classification beyond what is achievable with kinematic data alone?
• RQ2 How effectively can IMU-based kinematic data from a wearable neck-band be mapped and simulated within an OpenSim musculoskeletal model?
• RQ3 To what extent can machine learning models, particularly Random Forest, classify neck postures using combined kinematic and kinetic features?
• RQ4 Can the proposed system reliably detect and notify users of improper neck postures in real time to prevent musculoskeletal disorders?
• RQ5 What is the predictive performance of the system when using tendon force data from the Hyoid muscles as a kinetic input?

Key findings

• The Random Forest classifier achieved 100% accuracy in predicting nine distinct neck postures using kinematic data (acceleration and position) from the IMU.
• The same classifier achieved 100% accuracy in predicting neck postures using kinetic data derived from tendon forces of the Hyoid muscles in the OpenSim model.
• The integration of IMU kinematic data with OpenSim-simulated kinetic data significantly enhanced the reliability and precision of posture classification.
• The functional calibration of IMU data to the OpenSim model enabled accurate reconstruction of neck joint kinematics and muscle force dynamics.
• The system successfully demonstrated real-time potential for detecting improper neck postures through predictive analytics, supporting preventive healthcare applications.
• The results confirm that kinetic data from the Hyoid muscles provides a highly discriminative signal for posture classification, even when kinematic data alone shows limitations.